refactor: split tts.py into per-backend packages

1 files changed, 57 insertions, 0 deletions

diff --git a/app/converter/clients/speakers.py b/app/converter/clients/speakers.py
new file mode 100644
index 0000000..eecd52a
--- /dev/null
+++ b/app/converter/clients/speakers.py
@@ -0,0 +1,57 @@
+"""Qwen3-TTS built-in speaker names and their wire (display) forms."""
+
+from typing import Optional
+
+from .. import config
+
+# Built-in CustomVoice speaker names for the Qwen3-TTS family. Shared by the
+# qwen-tts demo backend (config.SPEAKER, the qwen setup/form) and the
+# audio.cpp audiocpp backend's CustomVoice entry (the Convert form's Speaker
+# picker). Entries are the canonical/config form; speaker_display_name()
+# maps them to the wire (display) form via SPEAKER_DISPLAY_NAMES below.
+QWEN3_TTS_SPEAKERS = ("Vivian", "Serena", "Uncle_Fu", "Dylan", "Eric",
+                      "Ryan", "Aiden", "Ono_Anna", "Sohee")
+
+# Canonical speaker names -> display names used by the qwen-tts demo.
+SPEAKER_DISPLAY_NAMES = {
+    "ryan": "Ryan",
+    "serena": "Serena",
+    "vivian": "Vivian",
+    "uncle_fu": "Uncle Fu",
+    "aiden": "Aiden",
+    "ono_anna": "Ono Anna",
+    "sohee": "Sohee",
+    "eric": "Eric",
+    "dylan": "Dylan",
+}
+
+
+def speaker_display_name_for(name: str) -> str:
+    """Return the wire (display) form of a Qwen3-TTS CustomVoice speaker NAME.
+
+    Accepts either the canonical/config form (e.g. "uncle_fu", "Uncle_Fu")
+    or the display form ("Uncle Fu"), case-insensitively; unknown names pass
+    through unchanged. Used by AudioCppTTSClient to normalize the --voice /
+    Speaker-picker value into what audiocpp_server expects in the request's
+    voice field.
+    """
+    return SPEAKER_DISPLAY_NAMES.get((name or "").lower(), name)
+
+
+def is_builtin_speaker(name: Optional[str]) -> bool:
+    """True when NAME is one of the Qwen3-TTS CustomVoice built-in speakers.
+
+    Matches case-insensitively across the canonical ("Uncle_Fu"), display
+    ("Uncle Fu") and shorthand ("uncle_fu") forms, so the --voice flag and
+    the Convert form's Speaker picker resolve to the same set.
+    """
+    if not name:
+        return False
+    norm = name.lower().replace("_", " ").replace("-", " ")
+    return any(norm == speaker.lower().replace("_", " ")
+               for speaker in QWEN3_TTS_SPEAKERS)
+
+
+def speaker_display_name() -> str:
+    """Return the display name for the configured custom speaker."""
+    return speaker_display_name_for(config.SPEAKER)